Julia High Performance - Second Edition

Using the profiler

The profiler lives within the Profile standard library package. So the first step in using the profiler is to import its namespace into the current session. You can do this using the using command:

julia> using Profile

This makes the @profile macro available. This measures and stores the performance profile of the expression supplied to it.

Do not profile the JIT:As with measuring the time of execution, remember to run your code at least once before attempting to profile it. Otherwise, you will end up profiling the Julia JIT compiler, rather than your code. If you see many instances of inference.jl in your profiler output, that means you are profiling the compiler instead of your code. If you see this, clear the profile ...